AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

How robust is unsupervised representation learning to distribution shift?

Yuge Shi, Imant Daunhawer, Julia E Vogt, Philip Torr, Amartya Sanyal

ICLR 2023 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Data-efficient and robust self-supervision · 6 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use SSL (avg of SimCLR, SimSiam, BYOL) instead of supervised learningrepresentation learner · MNIST-CIFAR CIFAR labelOOD accuracy: 51.5 -> 86.1Improved
use autoencoder family (AE, VAE, IWAE, beta-VAE) instead of supervised learningrepresentation learner · MNIST-CIFAR CIFAR labelOOD accuracy: 51.5 -> 79.9Improved
use SSL instead of supervised learning under extreme shift r_id=1representation learner · CdSprites shapeOOD accuracy: 44.0 -> 82.7Improved
use AE instead of supervised learning under extreme shift r_id=1representation learner · CdSprites shapeOOD accuracy: 44.0 -> 52.5Improved
use SSL instead of supervised learningrepresentation learner · Camelyon17OOD accuracy: 86.8 -> 89.8Improved
use SSL instead of supervised learningrepresentation learner · FMoW worst-groupOOD accuracy: 35.6 -> 29.6Hurt
use AE instead of supervised learningrepresentation learner · FMoW worst-groupOOD accuracy: 35.6 -> 22.8Hurt
use SSL instead of supervised learningrepresentation learner · FMoW worst-groupshift sensitivity: 37.7 -> 9.1Improved
use SSL instead of supervised learningrepresentation learner · MNIST-CIFAR CIFAR labelshift sensitivity: 47.6 -> 13.7Improved
retrain linear head on OOD data instead of ID data (ablation)SSL representations · MNIST-CIFAR CIFAR labelOOD accuracy (linear head bias b): 34.8Improved
retrain linear head on OOD data instead of ID data (ablation)supervised representations · MNIST-CIFAR CIFAR labelOOD accuracy (linear head bias b): 0.8No clear effect
retrain linear head on small OOD data (10% of OOD test) (ablation)AE and SSL · Camelyon17OOD accuracy (linear head bias b): 31.7 (AE), 8.8 (SSL)Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy